Emergency care in 59 low- and middle-income countries: a systematic review

1Ebola virus disease, cholera, armed conflict and natural disasters have recently strained systems for the provision of emergency care in low- and middle-income countries (LMICs). Expert groups have voiced concern about these systems’ critical lack of surge capacity and resilience. Even in non-crisis situations, small surveys, and anecdotal accounts hint at high volumes of critically-ill patients seeking emergency care in LMICs. This makes emergency care different from other health settings – including primary care – where doctors typically see only 8–10 ambulatory patients per day.

2In high-income countries, decades of advances in clinical science and care delivery have dramatically improved process efficiency and patient outcomes for a range of acute conditions.– Despite increasingly urgent calls to apply lessons learnt in high-income countries to LMICs,– a lack of data from the field has made it difficult to convince policy-makers to make major new investments in emergency care. Measuring the state of emergency care in LMICs is challenging, because care is delivered through a heterogeneous network of facilities and medical records are often incomplete, even for basic information such as patient identity and diagnosis.–

3Because of these challenges, studies of emergency care in LMICs have been limited to small, ad hoc efforts, in individual facilities, that were focused on individual acute diseases and conditions.– We systematically reviewed all available evidence on emergency care delivery to guide future research on – and improvements of – emergency health systems in LMICs.

4We did a systematic review (PROSPERO: CRD42014007617) – following PRISMA guidelines – to identify quantitative data on the delivery of emergency care to an undifferentiated patient population in all LMICs categorized as such in 2013. To increase capture, we also included the names of the autonomous or semi-autonomous geographical areas recognized by the World Bank and then disaggregated any relevant data obtained for such areas. For each country or subregion, we searched PubMed, CINAHL and World Health Organization (WHO) regional indices, using “emerg*” plus the country or area name as the search term. We wished to identify studies of emergency care, irrespective of location, patient complaint or provider specialty. We performed similar searches in Google Scholar but only searched within article titles. We also identified non-indexed journals that regularly published manuscripts on emergency care (available from the corresponding author) and screened every article in every issue of these journals manually. Searches were conducted between 12 August 2013 and 30 May 2014.

5We screened reports based on their titles and abstracts in English or French. The full-text potentially relevant articles were retrieved, irrespective of language or date of publication. Since the purpose of our review was to synthesize recent evidence on emergency care, the findings summarized below relate only to data published after 1989. A summary of our observations on data that were published before 1990 is available from the corresponding author. We retained studies describing the delivery of any emergency care in a health facility to adult or paediatric patients, irrespective of the presenting complaint or condition. For each retained article, we conducted backward and forward reference searches: we screened the references cited and, using Google Scholar, we also identified and screened publications that cited the article. We excluded studies that focused on specific conditions or subsets of emergency patients unless they also provided data on the overall population or facility. We also excluded studies that aggregated data from multiple facilities and general descriptions of the state of emergency care in a country. Despite the assistance of trained medical librarians, the full texts of some potentially relevant manuscripts could not be traced. In these cases, we used data from related abstracts or posters, when available.

6We presented the study protocol and early results at the 2013 African Federation for Emergency Medicine consensus conference. We made use of this presentation and our professional networks to request relevant unpublished data from clinicians in LMICs. Some clinicians, researchers and authors were not authorized to release data that allowed the study health facility or facilities to be identified. In these cases, we identified facilities only by their locations and ownership – i.e. academic, non-profit or for-profit.

7We extracted data on the characteristics of each study facility: country, urban or rural setting, bed count, annual patient volume, ownership and highest level of provider training. We considered a provider to be an emergency physician if reference was made to specialty postgraduate training, board certification or practice within an independent department of emergency medicine. We recorded details of the study population – i.e. age, sex, number of subjects included in analysis, number who arrived by ambulance – the sampling method and key patient outcomes. The latter included the inpatient admission and mortality within the emergency department, the percentages of patients recorded as brought in dead, or dead on arrival, and the length of time each patient stayed in the emergency department.

8We created a database containing aggregated study data. When multiple publications described a single facility, we merged them to create a single record that, for each variable of interest, contained the most recently published data available. We stratified facilities using World Bank regions and considered separately those facilities that only served paediatric populations. If data from a single facility were available disaggregated by age group, we summarized quantitative metrics for adult and paediatric patients separately. Full lists of the included studies and the data extracted and a full description of the study protocol are available from the corresponding author.

9We calculated summary statistics for all relevant metrics that were reported consistently across studies: bed count, annual patient volume, admission and mortality within the emergency department. We made an a priori decision not to perform a formal meta-analysis. Instead, our systematic analysis was meant to capture the distribution of metrics across populations – e.g. adult versus paediatric – and World Bank regions – e.g. Africa versus Asia – as well as global patterns. We thus present means – or medians with interquartile ranges (IQR) – disaggregated by country or region, as appropriate. Statistical analyses were performed using Stata/MP (StataCorp. LP, College Station, United States of America).

10Fig. 1 shows the results of our literature search. Of the 195 relevant published studies identified (Table 1; available at: http://www.who.int/bulletin/volumes/93/14/07-148338), 170 (87%) were descriptive reports on hospital-based emergency departments whereas the other 25 (13%) described the impact of an intervention. We obtained relevant unpublished data on a further 16 facilities. After combining multiple reports from the same facility and separating paediatric and adult data – for the three facilities with disaggregated data – we had data on 192 individual facilities in 59 countries. Of the 192 facilities, 107 (56%) were academically affiliated, 11 (6%) were in rural areas and 36 (19%) served paediatric patients exclusively; in the remaining 38, facility type could not be identified. Further information on the health facilities is available from the corresponding author.

11Table 2 presents the key metrics for the facilities. Median mortality within the emergency departments – of the 65 facilities that reported the relevant data – was 1.8% overall and higher in the 19 paediatric facilities (4.8%) than in the 46 adult or general facilities (0.7%). Across World Bank regions that we investigated, mortality was highest in sub-Saharan Africa (3.4%; IQR: 0.5–6.3%; n = 44), especially in east, central or west Africa (4.8%; IQR: 3.3–8.4%; n = 30). Paediatric facilities in sub-Saharan Africa had a median mortality of 5.1% (IQR: 3.5–11.1%; n = 15). Mortality in emergency facilities was also high in Latin America. Two facilities in Brazil were major contributors to this high rate, with mortality of 7.4% and 3.9%. These centres also reported long inpatient stays: one facility reported a median length of stay of three days, whereas the other reported that 21% of patients stayed in the emergency department for more than five days. Lengths of stay were only reported for 15 facilities and for these, the median value was 7.7 hours (IQR: 3.3–40.8). As mortality data were only available for nine of these 15 facilities, it was not possible to formally investigate the relationship between length of stay and mortality. The five sub-Saharan African facilities that recorded length of stay reported a median stay of 17 hours (IQR: 16.9–18.0). Additional data comparing mortality, patient volumes and admission are available from the corresponding author.

12Median annual patient volume was 30 021 (IQR: 10 296–60 000) among 173 facilities reporting these data. Volume was lower in the nine rural facilities (16 468; IQR: 3429–44 395) than in the 164 urban ones (31 000; IQR: 10 994–61 313). The 17 paediatric facilities in Sub-Saharan Africa had relatively low patient volumes with a median annual patient volume of 3129 (IQR: 2009–7479). The median inpatient admission was 20% (IQR: 10–43%; n = 78) and the median number of beds in the emergency department was 14 (IQR: 8–22; n = 60). The median age of patients attending non-paediatric facilities was 35 years (IQR: 6.9–41.0; n = 51) and a median of 55.7% (IQR: 50.0–59.2%; n = 93) were male. The corresponding values for paediatric facilities were 3.2 years (IQR: 2.8–3.4; n = 13) and 58.3% (IQR: 55.4–60.1%; n = 27), respectively.

13Table 3 summarizes the training of providers staffing the 102 facilities for which provider data were available. Care in 67 (66%) of these facilities was provided either by trainees or by physicians whose level of training was not specified. In only 29 (28%) of facilities were attending or consultant-level physicians available full-time; in 19 other facilities, physicians were only available in daytime hours. Eighteen facilities were staffed by specialty-trained emergency physicians, but in only four facilities were emergency physicians available at all times – one in the United Republic of Tanzania (unpublished observations, 2014), one in Pakistan and two in Nicaragua. One facility provided specialized emergency training to non-physician providers staffing the emergency department. In another facility, medical students practising alone were primarily responsible for providing emergency care during most of the day. Patients had to navigate through a wide range of options to obtain emergency care and financial factors played a major role in determining what kind of care they received (details available from the corresponding author).

14While only a small set of metrics on the delivery of emergency care were reported consistently across facilities, we were able to draw some conclusions on the state of emergency care in low-resource settings.

15First, large numbers of patients presented to health facilities seeking emergency care. While there was a wide range in annual patient volumes – from just 451 in a paediatric emergency department in Nigeria to 273 182 in a general emergency department in Turkey – they were approximately 10 times higher than the corresponding caseloads observed in primary care settings in sub-Saharan Africa and Asia.

16Second, patients seeking emergency care were generally young and free of chronic conditions. This is in contrast to the growing burden of elderly patients with multiple chronic conditions seen in the emergency departments of high-income countries. Therefore, interventions to decrease mortality and morbidity in emergency settings of LMICs could dramatically increase life-years saved and productivity.

17Third, the mortality recorded in emergency departments in LMICs was many times higher than generally reported in high-income countries.– A recent report on emergency departments in the USA documented a mean mortality within the departments of 0.04%.

18Fourth, most providers of emergency care in LMICs had no specialty training in emergency care. This observation was expected given the general shortages in human resources for health in most of these countries. Such shortages may be particularly pronounced in emergency settings, where the work is demanding and salaries are often poor. Most governments do not include emergency medicine in their medical education priorities.

19What implications do these results have for LMICs? We made a rough calculation for Nigeria, where we identified relevant studies in 21 facilities and mean annual patient volume of 3000 and 5–7% mortality. If we assume that the approximately 1000 teaching and general hospitals in the country have the same mean annual patient volume and mortality, then out of the 1.6 million deaths recorded annually in Nigeria an estimated 10–15% occur in emergency departments. This estimate – and the observation that most emergency departments in LMICs are run by providers with no speciality training in emergency care – illustrates the opportunity to improve emergency care in LMICs. It is likely that relatively simple interventions to facilitate triage and improve patient flow, communication and the supervision of junior providers (Box 1) could lead to reductions in the mortality associated with emergency care.–

20Local paramedics and lay first responders were trained to provide field care for trauma. After the intervention, the trauma mortality decreased from 40% to 15%.

21The paediatric clinic was physically restructured to streamline operations, clinical staff were trained in emergency care and triage and cooperation between the inpatient and outpatient services was improved. After the intervention, mortality within 24 hours of presentation decreased from 36% to 13%.

22A triage unit was established in the outpatient department and the emergency and intensive care units were combined. Clinical staff were trained in emergency care and triage, with experienced nursing and medical officers required to be present at all times. Equipment and record keeping were also enhanced. After the intervention, inpatient mortality decreased from 12% to 6%.

23The paediatric clinic allocated senior medical staff to supervise emergency care and implemented formal triage procedures, with an emphasis on early patient treatment and stabilization before transfer to the inpatient ward. Inpatient mortality within two days of admission decreased, from 5% to 4%.

24Our data illustrate the unique cost–benefit profile of investments in emergency care. Although disease and injury prevention are key functions of all health systems, acute health problems – e.g. myocardial infarction, sepsis and trauma – continue to occur in all countries. With the same amount of resources, it is likely that more lives could be saved in a paediatric emergency facility with mortality between 12% and 21%– than in paediatric primary-care clinics in similar settings – which generally see just a few critically-ill children per clinic per week (unpublished observations, 2015). There is thus a clear case for investing in emergency care in LMICs, to complement existing efforts to strengthen primary and preventive care.

25What is needed to strengthen emergency care in LMICs? First, a better understanding of the conditions that drive patients to seek such care is crucial. We documented high patient volumes and mortality but did not identify the diseases or the conditions that drive these metrics. While useful estimates of the burden of acute conditions may be produced in mathematical models, the setting of specific clinical and policy priorities remains difficult because of the scarcity of relevant data.

26Second, once we have a better understanding of the burden of acute disease, interventions known to be effective in high-income settings – e.g. trauma resuscitation training – must be adapted to LMICs and critically assessed. Some effective interventions to decrease mortality in emergencies (Box 1) may only require the improved use of existing system components, with the minimal input of new material resources. However, assessing the effectiveness of such interventions by rigorous experimental or quasi-experimental methods requires additional funding. Although before-and-after comparisons may be easier, they are also vulnerable to a range of biases.

27Third, international organizations must accelerate efforts to develop consensus on the essential components of systems for emergency care. Policy-makers who wish to assess their emergency systems and set priorities for development need technical guidance. WHO’s framework on systems of trauma care is one useful model for this broad agenda.

28Finally, improvements to emergency care in LMICs will require advances in data collection. The development of a minimum set of indicators for emergency care in LMICs would facilitate research and quality improvement. Several actors are improving platforms for data collection in LMICs. For example, the African Federation for Emergency Medicine is building consensus around a medical chart that has been purpose-designed to capture data for clinicians, administrators and researchers in LMICs. A novel data collection platform has been implemented for trauma care in a large teaching hospital in the United Republic of Tanzania, with promising early results. The systematic integration of routine data collection into care delivery settings should help ensure that interventions are – and remain – effective.

29The most important limitation of our study is the general paucity of data on emergency care. After screening over 40 000 published reports, we identified relevant data from only 192 facilities spread across 59 LMICs. For comparison, there are about 5000 emergency departments in the USA. The facilities we identified were largely urban and academic – as might be expected given that our search strategy relied mainly on published reports. Broader reporting biases may also have affected our results. For example, facilities with fewer resources may be relatively unlikely to collect and publish data and facilities with exceptionally high levels of mortality may be relatively unlikely to publish those levels. Thus, our results are likely to present an optimistic view of the state of emergency care in LMICs.

30Regional comparisons must be viewed with caution, given the geographical variation in facility characteristics and reporting practices. For example, emergency departments in which patients have exceptionally long lengths of stay will probably also have exceptionally high mortality – since patients who stay longer in the department are more likely to die in the department. Although a lack of relevant data prevented us from investigating this relationship, the median length of stay in our sample – albeit in the small number of facilities that reported lengths of stay – was only 7.7 hours. It therefore seems unlikely that prolonged stays alone could have accounted for the high levels of mortality that we observed.

31Other limitations were our search strategy, which relied on the presence of at least one word that began with “emerg” in the title, keywords or abstract of an article. While this made a difficult search problem tractable, it may also have excluded some relevant studies. Also, lack of data standardization across facilities and countries probably biased our results. For example, standardized measures of mortality – e.g. the percentage of patients that died with 24 hours of their presentation – were seldom reported, probably because of the difficulties of following-up patients after they leave the emergency department. The maximum age for a so-called paediatric patient also varied widely across studies, from five to 19 years.,

32Emergency facilities in LMICs serve a large, young patient population with high levels of critical illnesses and mortality. This suggests that emergency care should be a global health priority. The cost–benefit ratio for improvements in emergency care is likely to be highly favourable, given the high volume of patients for whom high-quality care could be the difference between life and death. There are likely to be substantial opportunities to improve care and impact outcomes, in ways that could be rigorously evaluated with manageable sample sizes.